Genetic algorithms and artificial neural networks method were adopted to .optimize the camber line shape of splitter cascade in a transonic axial fan. For a certain type of transonic axial fan, the isentropic effi-ciency was improved by 0.6% with the modification of splitter cascade rotor. Numerical simulation studies show that the curvature of the optimum splitter cascade with larger suction side is increased, the static pressure gradi-ent is reduced, the process of flow expansion on suction surface is controlled and the separation near hub is e-liminated by changing the shock wave structure. The efficiency near hub is improved by 1.3%.